ASSESSING THE PREDICTIVE VALIDITY OF THE SALZBERG SCALE DURING ACUTE CARE AND INPATIENT REHABILITATION. Rachelle Brick

Transcription

1 ASSESSING THE PREDICTIVE VALIDITY OF THE SALZBERG SCALE DURING ACUTE CARE AND INPATIENT REHABILITATION by Rachelle Brick Rehabilitation Sciences, University of Pittsburgh, 2014 Submitted to the Graduate Faculty of School of Health and Rehabilitation Sciences in partial fulfillment of the requirements for the degree of Bachelor of Philosophy University of Pittsburgh 2014

2 UNIVERSITY OF PITTSBURGH SCHOOL OF HEALTH AND REHABILITATION SCIENCES This thesis was presented by Rachelle S. Brick It was defended on March 28, 2014 and approved by Yi-Ting Tzen, Ph.D, Rehabilitation Science and Technology Patricia Karg, M.S., Rehabilitation Science and Technology Susan Garber, M.A., O.T.R, Baylor College of Medicine Thesis Director: David Brienza, Ph.D, Rehabilitation Science and Technology ii

3 Copyright by Rachelle S. Brick 2014 iii

4 ASSESSING THE PREDICTIVE VALIDITY OF THE SALZBERG SCALE IN ACUTE CARE AND INPATIENT REHABILITATION Rachelle S. Brick, B.Phil University of Pittsburgh, 2014 Pressure ulcers (PrU) are a leading secondary medical complication in the spinal cord injury (SCI) population. With over two hundred known risk factors, PrU prevention is extremely complex but can provide an astounding difference in a patient s recovery. Multiple risk assessment scales allow us to quantify risk across a broad range of populations, yet the literature provides little evidence that these tools are representative of PrU development in the SCI population. The Pressure Ulcer Assessment Scale for the Spinal Cord Injured (Salzberg Scale) is a risk assessment scale specific to the SCI population, composed of fifteen risk factors that divide PrU development risk into four categories. The objective of this thesis is to assess the predictive validity of the Salzberg Scale during acute care and inpatient rehabilitation following spinal cord injury. Data was extracted from a primary study on PrU outcomes for newly injured traumatic SCI patients in acute care and inpatient rehabilitation. A secondary analysis assigned subjects a raw Salzberg Scale score based on collected medical information and Salzberg Scale component definitions. The Salzberg scores were used to compute sensitivities, specificities, and accuracy of the scale with newly defined risk cut-off scores for acute hospitalization and inpatient rehabilitation. Sensitivity and specificity were calculated for the scale s ability to predict PrU ranging from two to 22 days after administration of the Salzberg Scale risk assessment tool. The use of iv

5 the scale in the acute care hospitalization to assess risk for PrU within 2-3 days showed the only strong predictive results with an area under the curve (AUC) of at the indicated cutoff score of 15. The sensitivity of 100.0% and a specificity of 75.0%, showed a more accurate prediction balance than the validation study conducted by Salzberg on a broader population sample. Overall, failure of the scale s predictive ability to predict a pressure ulcer over a longer time period suggests further studies must be completed in order for the scale be recommended for implementation in an inpatient rehabilitation setting. v

6 TABLE OF CONTENTS LIST OF EQUATIONS... XIII PREFACE... XV 1.0 INTRODUCTION... I 1.1 DISCUSSION OF PROBLEM... I 1.2 OBJECTIVES AND SPECIFIC AIMS... II 1.3 ORGANIZATION OF THE THESIS... III 2.0 REVIEW OF THE LITERATURE PRESSURE ULCER DEVELOPMENT IN INDIVIDUALS WITH SPINAL CORD INJURY (SCI) Definition Clinical Diagnosis and Staging Etiology Risk Factors Incidence and Prevalence Financial Impact ASSESSMENT SCALES TO ASSESS PRESSURE ULCER RISK SCALES FOR GENERAL POPULATION The Braden Scale, Norton Scale, and Waterlow... 8 vi

7 3.1.2 Validity for SCI Patients SCALE FOR SCI POPULATION Development of Scales Pressure Ulcer Risk Assessment for the Spinal Cord Injured Risk Category Breakdown Previous Use of Scale RESEARCH DESIGN AND METHODS RESEARCH DESIGN SUBJECT INCLUSION AND EXCLUSION CRITERIA Primary Study Subject Inclusion and Exclusion Criteria Secondary Study Subject Inclusion and Exclusion Criteria DATA COLLECTION DATA ANALYSES Box Plots Sensitivity and Specificity Predictive Values ROC Curve RESULTS DESCIPTIVE STATISTICS Exploratory Analysis of Next Visit Predictions Groups Exploratory Analysis of the Multiple Visits Prediction Groups PSYCHOMETRIC PROPERTIES EVALUATION Next Visit Prediction Group vii

8 Total Population Acute Hospitalization Only Inpatient Rehabilitation Only Multiple Visits Prediction Group Total Population Acute Hospitalization Only Inpatient Rehabilitation Only DISCUSSION CONCLUSIONS, LIMITATIONS, FUTURE WORK BIBLIOGRAPHY viii

9 LIST OF TABLES Table NPUAP Staging System... 3 Table 2 Salzberg Scale Risk Categories Table 3. Data Collected for RERC-SCI Table 4. Secondary Analysis Data Extracted Table 5 Number of Observation Pairing Contributed Per Subject in Next Visit Prediction Group Table 6 Number of Observation Pairing Contributed Per Subject in Multiple Visits Prediction Group Table 7 Summary of Sensitivity and Specificity Calculation Table 8 Clinical Results Organizer for Sensitivity and Specificity Table 9 Summary of Predictive Validity Terminology Table 10 Demographics of Next Visit Prediction Populations Table 11 Demographics of Multiple Visits Prediction Populations Table 12 Observation Pairings of the Next Visit Prediction Group Table 13 Observation Pairings of the Multiple Visits Prediction Group Table 14 Total Population Box Plot Results of Next Visit Prediction Group Table 15 Acute Care Hospitalization Group Box Plot Results for Next Visit Prediction Group. 39 Table 16 Inpatient Hospitalization Box Plot Results for Next Visit Prediction Group Table 17 Total Population Box Plot Results for Multiple Visits Prediction Group ix

10 Table 18 Acute Care Hospitalization Box Plot Results for Multiple Visits Prediction Group Table 19 Inpatient Rehabilitation Box Plot Results for Multiple Visits Prediction Group Table 20 Tests of Salzberg Scale Cutoff Points in the Total Group Table 21 Statistical Calculation Table for Data Set in Acute Phase Table 22 Statistical Calculation Table for the Data Set in Inpatient Phase Table 23 Statistical Calculation Table for the Data Set in Total Group Table 24 Statistical Calculation Table for Data Set in Acute Phase Table 25 Statistical Calculation Table for the Data Set in Inpatient Phase x

11 LIST OF FIGURES Figure 1 Data Collection Process Figure 2 Subject Selection Flow Chart Figure 3 Generic Box Plot Characteristics Figure 4 Total Population Exploratory Comparison of Next Visit Prediction Group Figure 5 Acute Care Hospitalization Exploratory Comparison in Next Visit Prediction Group.. 38 Figure 6 Inpatient Rehabilitation Exploratory Comparison for Next Visit Prediction Group Figure 7 Total Population Exploratory Comparison for Multiple Visits Prediction Group Figure 8 Acute Care Hospitalization Exploratory Comparison for Multiple Visits Prediction Group Figure 9 Inpatient Rehabilitation Exploratory Comparison for Multiple Visits Prediction Group Figure 10 Cut off Plot for Total Population Figure 11 ROC of Salzberg Scale for Total Figure 12 Cut off Plot for Acute Hospitalization Figure 13 ROC Curve of Salzberg Scale for Acute Phase Figure 14 Cut off Plot for Inpatient Rehabilitation Figure 15 ROC of Salzberg Scale for Inpatient Rehabilitation Figure 16 Sensitivity vs. Specificity of Total Group Figure 17 Total Group ROC Curve xi

12 Figure 18 Sensitivity vs. Specificity of Acute Care Hospitalization Group Figure 19 Acute Care Hospitalization Only ROC Curve Figure 20 Sensitivity vs. Specificity of Inpatient Rehabilitation Group Figure 21 Inpatient Rehabilitation Group Only ROC Curve xii

13 LIST OF EQUATIONS Equation 1 Sensitivity Calculation Equation 2 Specificity Calculation Equation 3 Positive Predictive Value Equation 4 Negative Predictive Value xiii

14 xiv

15 PREFACE Multiple people have helped me through the entire Bachelor of Philosophy process, in which I am truly appreciative of their guidance. My sincerest thanks to: Dr. David Brienza, my faculty advisor, for answering countless questions, providing me with a strong research foundation, and believing in me throughout this nearly three year experience Shilpa, my incredible PhD Student Mentor who showed me that research can be a life changing career. I am so lucky to be able to graduate with you and see you become a Ph.D. Yi-Ting Tzen and Patricia Karg for your active role, guidance, and peer review Susan Garber for your continuous advice, encouragement, communication from a distance and expertise Karen Greenwald and Mary Beth Kusturiss for your countless hours of deciphering medical records and patient data with me Dr. Ellen Cohn, Amy Evans, and the School of Health and Rehabilitation Sciences for placing me in Dr. Brienza s lab and presenting the BPhil opportunity Sondra Balouris and Pamela Toto for sparking my interests in research and dedication to the rehabilitation sciences field. The entire faculty and staff of the RST Bakery Square Laboratory, especially Debby Keelan for all of your help, smiles, and support during this process. The University Honors College, specifically The Brackenridge Fellowship, and Chancellor s Undergraduate Research Fellowship, for funding my research Shanshan Tu for your abundance of knowledge in statistics and guidance through the University of Pittsburgh Statistical Consulting Center My family and friends for providing endless encouragement and support xv

16 1.0 INTRODUCTION 1.1 DISCUSSION OF PROBLEM Pressure ulcers (PrU) are a leading secondary medical complication in the spinal cord injury (SCI) population. 51, 65 Pressure ulcer development has been known to interfere with the physical, psychological, and social aspects of well-being along with impacting overall quality of life. 48 With over 200 known risk factors and treatment costing about $1.3 billion annually, pressure ulcer prevention is extremely complex but can provide an astounding difference in a patient s recovery and save money for the patient and hospital system. 19 More than thirty-eight scales allow us to quantify risk across a broad range of populations, yet the literature provides little evidence that these tools are representative of pressure ulcer development in the SCI population. 7, 19, 51 Unlike the most popular and well researched scales such as the Braden Scale, the Norton Scale and the Waterlow Scale, the Pressure Ulcer Assessment Scale for Spinal Cord Injured (referred to as Salzberg Scale or SCIPUS) is a risk assessment scale created specifically to identify risk of pressure ulcer development in the SCI population. Proper identification of individuals with SCI at risk for PrU development and initiation of appropriate preventative practices is necessary to lower the PrU incidence and prevalence. 7 Evaluation of risk assessment scales and other general prevention measures have been flagged as

17 crucial by various health reviews in the United States. Healthy People 2010 lists reduction of pressure ulcer incidence as an a main objective for healthcare providers. 71 In addition, more research on preventative tools is urgent in accordance to the Deficit Reduction Act of As of 2008, hospitals are no longer to receive compensation or reimbursement for care related conditions such as pressure ulcers that occur during incident hospitalization as they were deemed a preventable condition. 32 Overall, if risk assessment and prevention are not incorporated in healthcare practice, pressure ulcer development will continue to be a burden on the lives of those with SCI in various physical, psychological, as well as economical ways. Defining the risk cut off scores of the Salzberg Scale is essential in order to properly assess the already high-risk SCI population. Limited use of the Pressure Ulcer Assessment Scale for the Spinal Cord Injured suggests positive findings for the general SCI population, yet the tool has not been analyzed for the acute hospitalization and inpatient rehabilitation of new, traumatic SCI individuals. The objective is to evaluate the psychometric properties of the Salzberg scale in order to provide evidence predictive validity in new, traumatic SCI individuals. 1.2 OBJECTIVES AND SPECIFIC AIMS The objective of this thesis was to assess the predictive validity of the Salzberg Scale during acute care and inpatient rehabilitation following spinal cord injury. This retrospective secondary analysis covered three main aims. Each aim was accomplished using the Rehabilitation Engineering Research Center (RERC) on Spinal Cord ii

18 Injury clinical data collected on new, traumatic spinal cord injury subjects from acute care and inpatient rehabilitation settings. Collected from 2008 through 2012, the data set included detailed objective clinical tests including diagnoses and medical comorbidities, pain and depression scores, ambulatory and medication descriptions, as well as Braden Scale scores for each subject assessment date. Through data outcome compilation, raw Salzberg Scale scores were calculated and used for analysis of psychometric properties of the scale. The initial aim verified the generalized statement that the Salzberg Scale predicts pressure ulcer formation in the SCI population. The descriptive analysis assessed the difference in scores between those with and without pressure ulcer formation. The second aim evaluated risk cut-off points within the RERC study population. New sensitivities and specificities were calculated in order to define potential Risk Cut-Off Point associated with new, traumatic SCI. Each population will have its own specific risk scores associated with specific characteristics of its population demographics and injury severity. This is the basis of newly calculated risk cut-off point establishment. Thirdly, the new sensitivities and specificities were used to evaluate the accuracy and power of the Salzberg Scale scoring using an ROC Curve. 1.3 ORGANIZATION OF THE THESIS Six chapters follow the introduction. Chapter 2 is a review of the literature on pressure ulcer research focusing on risk factors for the development of pressure ulcers in the SCI and other populations. Chapter 3 is a review of PrU prevention techniques, specifically the use of validated risk assessment. Chapter 4 contains a description of the research design and methods iii

19 used in this study. Chapter 5 provides the results. Chapter 6 is a discussion of the results. Lastly, Chapter 7 summarizes the work, explains limitations, and gives future directions of the research. iv

20 2.0 REVIEW OF THE LITERATURE 2.1 PRESSURE ULCER DEVELOPMENT IN INDIVIDUALS WITH SPINAL CORD INJURY (SCI) Definition For this study, the definition of a pressure ulcer is defined by the National Pressure Ulcer Advisory Panel. Based on the 2012 revision, pressure ulceration is defined by the following: A pressure ulcer is a localized injury to the skin and/or underlying tissue usually over a bony prominence, as a result of pressure, or pressure in combination with shear. A number of contributing or confounding factors are also associated with pressure ulcers; the significance of these factors is yet to be elucidated. 56 The majority of pressure ulcers are found below the waist, specifically the skin overlying the bony prominences such as sacrum, trochanters, ischium, and heels, but can be found anywhere on the body where pressure and compressive forces are maintained for a sufficient period of time. 9, 49, 76 PrU are also known as decubitus ulcers, pressure sores, and/or bed sores. 9 1

21 2.1.2 Clinical Diagnosis and Staging Staging of PrUs is completed after the clinical diagnosis and identification of PrU development. This can be completed though visual inspection and/or palpation. 37 Various systems have been developed for the classification of pressure ulcers. Consistently recognized by the Wound Ostomy and Continence Nurses Society (WOCN), the National Pressure Ulcer Advisory Panel (NPUAP) staging system is a favored set of guideline for pressure ulceration diagnosis. 2, 22 Created in 1989, the original NPUAP staging system was composed of a four-stage system similar to the earliest systems created by Guttmann in Based on the growing popularity of the NPUAP Staging System, the European Pressure Ulcer Advisory Panel (EPUAP) also developed a similar system in The EPUAP Staging System was not adopted for diagnostic use in the United States since inter-observer reliability was low and there is still a need for clarity 1, 11 within the staging definitions. Following the creation of the NPUAP Staging System, factors such as accuracy, consistency, and inter-rater reliability of the staging system were investigated. Of the four stages, nurses exhibited uncertainty differentiating between Stage II and Stage III pressure ulcers. Diagnostic inaccuracies led to the reevaluation of the NPUAP Staging System. 13 At the 2005 Consensus Conference, NPUAP acknowledged the idea that suspected deep tissue injury (sdti) as well as unstageable categories were etiologies of pressure ulcer formation yet they could not be accurately described by the existing staging system. 14 Through solicitation of current facilities using NPUAP guidelines, the NPUAP presented a newly drafted scale for participants to comment on the qualities of clarity, succinctness, accuracy, and discrimination for each definition. 13 Refinement of the updated staging system was solidified at the 2007 NPUAP Consensus Conference. 29 The most recent revision (2012) was adopted and is currently used 2

22 today. The current NPUAP staging definitions are represented in Table Today, the NPUAP and EPUAP have a unified recommendation and definition system that is referenced often. 35 Table NPUAP Staging System Stage Suspected Deep Tissue Injury (DTI) Unstageable 2012 NPUAP Staging System Definitions Intact skin with non-blanchable redness of a localized area usually over a bony prominence. Darkly pigmented skin may not have visible blanching; its color may differ from the surrounding area. The area may be painful, firm, soft, warmer or cooler as compared to adjacent tissue. Category I may be difficult to detect in individuals with dark skin tones. May indicated at risk persons (a heralding sign of risk) Partial thickness loss of dermis presenting as a shallow open ulcer with a red pink wound bed, without slough. May also present as an intact or open/ruptured serum-filled filled blister. Presents as a shiny or dry shallow ulcer without slough or bruising.* This stage should not be used to describe skin tears, tape burns, incontinence associated with dermatitis, maceration or excoriation. *Bruising indicates deep tissue injury. Full thickness tissue loss. Subcutaneous fat may be visible but bone, tendon or muscles are not exposed. Slough may be present but does not obscure the depth of tissue loss exposed. Slough may be present but does not obscure the depth of tissue loss. May include undermining and tunneling. The depth of a Category/Stage III pressure ulcer varies by anatomical location. Full thickness tissue loss with exposed bone, tendon, or muscle. Slough or eschar may be present. Often includes undermining and tunneling. The depth of a Category/Stage IV pressure ulcer varies by anatomical location. Purple or maroon localized are of discolored intact skin or bloodfilled blister due to damage of underlying soft tissue from pressure and/or shear. The area may be preceded by tissue that is painful, firm, mushy, boggy, warmer or cooler as compared to adjacent tissue. Deep tissue injury may be difficult to detect in individuals with dark skin tones. Full thickness tissue loss in which the based of the ulcer is covered by slough (yellow, tan, gray, green, or brown) and/or eschar (tan, brown, or black) in the wound bed. Until enough slough and/or eschar are removed to expose the base of the wound, the true depth cannot be determined; but it will be either a Category/Stage III or IV. 3

23 2.1.3 Etiology Though the exact etiology of pressure ulceration is not fully understood, researchers have hypothesized pressure ulcer origins through four pathophysiological pathways. 75 The first of the hypotheses is based on localized ischemia caused by capillary occlusion reasoned by pressure loading. 23, 27 Ischemia, the insufficient supply of blood to an organ, prevents oxygen, essential nutrients and metabolites from being carried to and away from cells. This, in turn, causes the buildup of various cell waste products leading to the degradation and death of localized cells. 53,60 Another explanation refers to reperfusion injury theory. Reperfusion injury is an injury that results from the accumulation of inflammatory substances in the body as blood is reintroduced to an area previously affected by ischemia. 38 The third hypothesis of pressure ulcer encompasses the impaired lymphatic function in the body which causes accumulation of metabolic waste products and enzymes exhibiting similar responses as ischemia in a previous hypothesis. 50 Shear force may increase of PrU development. 43 Shear forces are parallel to the skin surface. It has the ability to damage blood vessels and compromise the blood supply leading to ischemia, necrosis, and cellular death. 49, 53 An example of shear force occurs when a person is in a hospital bed. When the back of the mattress is elevated, the weight of the upper body slides downward toward the bottom of the hospital bed. There is an opposing, parallel force that causes deformation. Though previously suggested that the frictional force may lead to the development of PrU, the 2009 International NPUAP-EPUAP Pressure Ulcer Prevention and Treatment Clinical Practice Guideline eliminated the frictional force as a factor. Despite friction resisting the shearing force, friction relates mainly to skin injuries that are not considered to be a PrU. Friction that causes shear strain in the tissue potentially leads to the increase in tissue breakdown, 4

24 which leads to PrU development. The sole force of friction cannot allow an injury to be categorized as a PrU. 6 Overall, there are two popular theories describing the mechanisms of pressure ulcer development and progression. The first theory states that pressure ulcers form deep within the muscle tissue, close to the bone, and move outward toward the superficial layers of the skin. This is known as the deep tissue injury theory. 40 Deep tissue injury arises when the muscle layers adjacent to the bone endure sustained loading. They are not necessarily visible until the unrelieved, ischemic injury and local necrosis reach the outer layers of the skin. 49, 53, 61, 73 The second, and lesser popular model, is the top to bottom model. This model states that pressure ulcer formation begins from the superficial skin destruction at the dermis through forces such as 15, 61 friction, and the injury proceeds inward toward the deeper tissue Risk Factors Due to the drastically altered physiology of a SCI patient immediately after injury, pressure ulcer formation is a leading health care problem. 41 Major factors associated with pressure ulcer development are level of activity, ambulation and level of mobility, incontinence, as well as severity of injury. 7, 10, 66 There are over 200 noted risk factors for pressure ulcer development in the SCI population. 19 The fifteen risk factors chosen by Salzberg will be discussed thoroughly in the third chapter of this thesis. These factors include level of activity, level of mobility, completeness of SCI, urinary incontinence or moisture, level of spasticity, age (years), tobacco usage, past or current pulmonary disease, past or current cardiac disease, glucose levels, past or current renal disease, impaired cognitive function, health care setting of pressure ulcer development, current albumin levels, and current hematocrit levels. 67 Each of these fifteen risk 5

25 factors designated in the Pressure Ulcer Risk Assessment Scale for the Spinal Cord Injured will be discussed in detail on later pages Incidence and Prevalence PrU incidence and prevalence varies by population and healthcare setting. In 2005 Chen et al conducted a study noted in the review using secondary complication data from the National Spinal Cord Injury Database. Chen et al. identified that the risk of developing a PrU was consistent over the first ten years post-injury (prevalence rate of 11.5% to 14.3% for Stage II or higher PrU) while risk eventually increased 15 years post injury (prevalence rate of 21.0% Stage II or higher PrU). 20 Pagliaccci et al found that 26.9% of 684 people with newly, traumatic SCI admitted to one of 32 rehabilitation centers in Italy presented with one or more PrU as well. 54 It has been recommended that the development of innovative strategies for prevention of PrU is needed to lower the high incidence and prevalence of PrU. 22 Incidence and prevalence of pressure ulcer development has been difficult to collect based on various pressure ulcer classification systems, inconsistent data collection and identification from multiple sources, as well as studies evaluating prospectively versus retrospectively. 31 According to data collected from the Model Spinal Cord Injury Systems (MSCIS) one third of individuals with SCI will develop at least one PrU in their initial acute or inpatient rehabilitation care. Thereafter, between 15%-33% of persons with SCI will develop pressure ulcers once in community settings. 41 With a range of 232,000 to 316,000 individuals with SCI in the United States, up to 85% will develop a pressure ulcer at some point during their life. 47, 62, 72 In addition, PrU-related complications cause up to 60,000 deaths a year, 7%-8% of these involving those with persons with SCI. 44 6

26 2.1.6 Financial Impact Though costs of treating pressure ulcers vary based on severity of the ulcer, each year, the United States spends more than $1.3 billion annually on treatment costs. 19, 43, 65 The highest estimation is upwards of $5 billion for treatment and management of pressure ulcers. 44, 52 In 2011, the cost of healthcare and living expenses of SCI patients ranged from $1,461,255 in the case of a minor injury all the way to the exorbitant $4,373,912 in the case of high level tetraplegia (C1-C4). 72 Out of these estimated healthcare and living expenses, 25% are attributed to treatment of pressure ulcers. 19 Partial thickness pressure ulcers (Stage I or Stage II) are estimated to cost $125-$451 per incidence. For more severe pressure ulcers, full thickness pressure ulcers (Stage III and Stage IV and sdti) cost an estimated $14,000-$25,000 per ulcer to treat. 58, 80 The cost of the pressure ulcer is higher with a more severe stage based on longer healing times and increased likelihood for complications. 46, 51 The magnitude of pressure ulcer development has been linked to increased hospital stays of approximately four days as well as prolonged nursing care time by nearly fifty percent. 30, 80 As noted, pressure ulcer development comes at a tremendous personal and societal cost. 45, 51 Implementation of preventative measures may decrease pressure ulcer 12, 25, 26, 34, 51 incidence and in turn lower its financial impact. 7

27 3.0 ASSESSMENT SCALES TO ASSESS PRESSURE ULCER RISK 3.1 SCALES FOR GENERAL POPULATION The Braden Scale, Norton Scale, and Waterlow Various risk assessment scales have been created in order to aid clinicians in the proper implementation of preventative measures for patients at risk for PrU development. Whether in a nursing home setting or a surgical unit, risk assessment scales have been made to be general or modified to a specific population at risk for developing PrU. 16, 59 It has been concluded that informal risk assessment such as skin integrity examination of a patient cannot take the place of a formal risk assessment. 7 Three of the most common PrU risk assessment scales are the Braden Scale, Norton Scale and the Waterlow Scale. The first risk assessment scale created to evaluate PrU development risk was the Norton Scale in Developed for the geriatric population, it is composed of five items: physical condition, mental state, activity, mobility, and incontinence. Each item is rated from 1 (very bad) to 4 (very good) with a maximum score of 20 points. 57 Scrutinized for its validity, patients with a cutoff score of fifteen or sixteen points or higher are considered at risk for pressure ulcer development. 30 Newer scales have been modeled off of the Norton Scale by redefining risk factors as described by recent research studies. 26 8

28 The second scale created and modeled after the Norton scale is the Waterlow Scale. Known to be more accurate at predicting pressure ulcers in orthopedic patients as compared to Braden Scale and Norton, the Waterlow Scale consists of eight items: build/weight for height, visual assessment of the skin in the area at risk, sex and age, continence, mobility, appetite, medications, and special risk factors. Unique to the scale is its weighted risk factors that vary in 30, 57 score distribution depending on the risk indicator contribution to PrU development. Receiving a score of 16 or above, defined cutoff point for at-risk patients in clinical studies predicts PrU development. This means that patients are considered at higher risk when there is an increase in the sum of the eight categories. The scale has patient risk broken down into 3, 78 categories of at risk, high risk, and very high risk. The third and most popular scale is the Braden Scale. Created it 1987, it has been concluded to be the best risk assessment scale due to its extensive reliability and validity testing to date. 55, 69 Composed of six subscales, the Braden Scale incorporates sensory perception, activity, mobility, moisture, nutrition, and friction/shear. Each subcategory is rated from 1 to 4 points with friction/shear rated only from 1 to 3 points. Unlike other scales, a score lower than points subjects a patient to be at higher risk for PrU development. Since its introduction, various cut-off scores have been recommended based on the specific population of patients. Studies validating the use of this scale have had sensitivities ranging from 79% to 100% over a variety of treatment settings. 43 The Braden scale has multiple strengths. Whether it be its high utility in clinical or research settings, existing validity and reliability evidence, or inclusion of the latest indicative factors of PrU risk, the Braden Scale has had consistent implementation since its creation. 39 9

29 Between these three scales, it is very difficult to compare the predictive validity as many validation studies, characteristic research methods, and outcome measures are evaluated in each study. 8 In some studies, the Waterlow and Braden are represented as having the best sensitivities whereas the Norton Scale has the best specificity. 8, 57 Contrasting these studies, there are articles which state that the Braden Scale results has superior predictive validity based on prospective versus retrospective studies. 24 Overall, all risk assessment instruments identified more patients at risk of PrU development over clinical judgment alone Validity for SCI Patients Though the Norton Scale, Waterlow Scale, and Braden Scale have been validated for use in hospital settings and nursing homes, they lack application and validation of specific populations such as the SCI population. 51 The predictive value of existing risk assessment tools are imprecise for the SCI population as they were originally designed for the general medical population. 33 WB Mortenson et al. completed a systematic review of risk assessment scales for pressure ulcer development using the SCI population. 51 Of the seven scales incorporated into the review, the Norton Scale, Waterlow Scale, and Braden Scale were evaluated. Based on previous studies regarding reliability, validity, administrator burden and respondent burden, each of these three scales had poor to adequate predictive validities with the spinal cord injury population. Being that the spinal cord injury population is initially at high risk for PrU development, it is crucial that a risk assessment scale identify the specific risk of this population. 36 As stated so precisely in an analysis describing usability of surveys on various population, [G]eneralizability Theory indicates that [previous] results cannot be applied to 10

30 individuals with SCI, which undermines confidence in these scales to reproduce stable results over time with this population. 21 This means that there were such a small subpopulation of SCI subjects included in the various studies assessing the three scales, that the overall predictive validity could not be attributed to the SCI subpopulation s contribution. Lack of content validity causes institutions to assume the scale s application to the population. Given that the Braden Scale and Waterlow performed similarly and closest to adequate scoring based on psychometric property evaluation, these scales are the best tools that are currently available despite benefitting from additional testing within the SCI population. 51 Included in the systematic review was the Salzberg Scale. This was the only scale made specifically for the risk assessment of the SCI population. Despite its higher sensitivity (74.7%) and specificity (56.6%) scores, it could not be deemed the best assessment tool since has not been validated. Mortenson et al. advised that the Salzberg Scale undergo further psychometric testing in order for recommendation. Lastly, more prospective studies that allow head-to-head comparison of these risk assessment scales would represent a stronger method to evaluate concurrent and construct validity for proper direct recommendation of a SCI specific risk 28, 51 assessment tool. 11

31 3.2 SCALE FOR SCI POPULATION Development of Scales With over 250,000 individuals with SCI in the United States, currently no method exists to accurately identify which of these individuals will have a higher risk for developing pressure ulcers. 67 This idea sparked the curiosity of colleagues Dr. Andrew Salzberg and Daniel Byrne. Developing a research interest in the early 1990s, Salzberg and Byrne set out to perform a systematic review of literature regarding the many risk factors associated with pressure ulcer development for people with SCI. 19, 52, 65 the Norton Scale, the Gosnell Scale, and the Braden Scale were addressed specifically for their lack of content validity in the systematic review. 19 In addition to these three scales, other papers have cited that the Waterlow Scale also lacked content 55, 79 and concurrent validity. Certain criticism pertaining to these scales revolve around their vaguely defined risk factors, lack of weighting of risk factor for emphasis, and better prediction based on medical practitioner intuition. 64, 77 In addition, inadequate number of individuals with SCI in sample sizes and various confounding variables make justifying the results of formerly created scales to the SCI population difficult. The results of the systematic review led to a reduction of the 200 known risk factors to a succinct list of fifteen. These fifteen tailored risk factors to the SCI population were incorporated in the Pressure Ulcer Risk Assessment for the Spinal Cord Injured. 12

32 3.2.2 Pressure Ulcer Risk Assessment for the Spinal Cord Injured The first study on the Pressure Ulcer Risk Assessment for the Spinal Cord Injured, commonly referred to as the Salzberg Scale was published in Salzberg and Byrne et al included the previously established fifteen risk factors that correlate to pressure ulcer development in the SCI population. Stated in their original study, these fifteen factors were included based on four main criteria. The fifteen risk factors must have some form of statistical association with pressure ulcer development in the SCI population. In addition, each risk factor included in the scale had a plausible biological mechanism, literature support as well as relate to the improved prediction of pressure ulcer development. 67 The fifteen risk factors included in the study were level of activity, degree of mobility, completeness of SCI, urinary incontinence, diagnosis of autonomic dysreflexia, age, comorbidities such as those pertaining to cardiac, pulmonary, and renal pathophysiology, level of cognition, diagnosis of diabetes, history of cigarette smoking, residency, and diagnosis of hypoalbuminemia and anemia. Each risk factor is weighted differently ranging from a value of 0, 1, 2, 3, or 4 points based on their significance and relevance to pressure ulcer development. Those with the largest score determination are level of activity, level of mobility, age, history of tobacco use, existing pulmonary disease, and location of residence. These all have a possibility of contributing two or more points to the total score at the time of assessment. To determine which score should have been assigned to the subject or patient, Salzberg provided an objective, easily defined list of operational definitions. Each category must be filled out to create a total score. Total scores range from 0 to 25 points. The lower the compiled score, the lower the level of risk a person will have for pressure ulcer development. 13

33 3.2.3 Risk Category Breakdown In the process of developing the Salzberg Scale, Salzberg and colleagues established four categories to define the amount of risk. The four categories can be seen in Table 2 below. Table 2 Salzberg Scale Risk Categories Risk Category Point Value/Range LOW 0-2 MODERATE 3-5 HIGH 6-8 VERY HIGH 9-25 These categories were based on initial sensitivity and specificity analysis of 219 subjects. Their data indicated that patients with a score greater than 6 had the highest balance of sensitivity and specificity, demonstrated by the intersection of the sensitivity and specificity lines. 67 As recommended by Salzberg, Braden, as well as W.B. Mortenson, each specific setting should test baseline risk categories on the population they study or treat. This way, the scale is personalized and specific for the appropriate settings use. It is essential that the tool be tested on the intended population prior to implementation Previous Use of Scale The Salzberg Scale has been used in a limited amount of studies. 34, 51, 67, 68 Its usage varies from its preliminary psychometric testing to additional medical record support in larger interventional studies. 10 Use of this scale is found in trace studies due to its lack of validity and popularity as a promising measurement tool. In 2007, W.B. Mortenson et al. composed a systematic review of risk assessment scales for pressure ulcer development in individuals with SCI. 51 Information on 14

34 the Salzberg Scale as well as six other risk assessment scales was evaluated based on published results of each scales reliability, validity, respondent burden, and administrative burden. Of all tools, the Salzberg Scale had the most items to assess with its unique categories of autonomic dysreflexia and living setting. It was identified that the Salzberg Scale had the best sensitivity and specificity of all measures tested during acute care hospitalization as well. It also had the highest construct validity. What diminishes the scale s high sensitivity and construct validity is its only application in retrospective studies and limited published results. 51 Despite the Salzberg Scale being customized for the assessment of individuals with SCI, there is still the criticism of minimal reliability data and general population use in other studies than its founding study. It has yet to be recommended until further psychometric evaluation is completed. 33 Additionally, the predictive value of the Salzberg Scale in various settings (e.g., community or hospital) requires investigation as well. 15

35 4.0 RESEARCH DESIGN AND METHODS Data collection was made possible by the Rehabilitation Engineering Research Center on Spinal Cord Injury (RERC-SCI). The RERC-SCI assists persons with SCI through research and development of technologies that can be incorporated into the improvement of rehabilitation and reintegration of persons into society. Made up of developmental and research projects, the clinical core of this particular study was designed to enroll new, traumatic SCI patients for the collection of demographic information, medical information, pressure ulcer outcomes, and urine and blood samples during the acute, inpatient, and outpatient stages of rehabilitation. For a complete listing of variables collected, see Table 3. Informed consent reviewed and approved by the University of Pittsburgh Institutional Review Board (IRB) was obtained for both the RERC-SCI study and amended for the retrospective secondary analysis. Patients enrolled were charted at the University of Pittsburgh Medical Center (UPMC) facilities and after discharge. The Clinical Core of RERC-SCI collaborated with UPMC and the UPMC Model System on SCI to recruit, screen, and enroll subjects with new, traumatic SCI. All personnel in the Clinical Core team were qualified to collect data and perform proposed responsibilities. Data was taken from each enrolled patient three times a week during acute care hospitalization, once a week during inpatient rehabilitation, and annually after discharge in outpatient care. Outpatient follow-up consisted of six months, twelve months, and yearly time points until the conclusion of the study. All evaluations 16

36 continued thru the termination of the study. The RERC-SCI was funded by the National Institute on Disability Rehabilitation Research grant, #H133E Table 3. Data Collected for RERC-SCI Variables Categories Demographics Injury Medical Information Pressure Ulcer Variable Specifics Admission Date, Consent Date, Date of Injury, Age, Sex, Marital Status, Ethnicity, Height, Weight, Past Medical History (PMH) of Smoking Cause of Injury; Level of Injury: Cervical, thoracic, Lumbar, Sacral, Coccyx, Other; Bilateral vs. left and right; Complete vs. Incomplete; American Spinal Injury Classification (ASIA) score, Lower Extremity Score Injury Severity Score (ISS) Alcohol Use, Tobacco Use, Musculoskeletal Pain, PHQ Depression Scale, Medical Comorbidities (cardiovascular, hematopoietic, respiratory, ENT, gastrointestinal, rheumatologic, musculoskeletal, neurologic, endometric/metabolic, immunological, psychiatric, malignancy, substance abuse, UTI, integumentary), Number of Pressure Ulcers, Braden Risk Score, Bladder and Bowel Management, Medications (NSAIDs, Steroids, Antibiotics, and Others), Ambulation Status Size, Shape, Severity (based on NPUAP Staging System), Progression of the PU 4.1 RESEARCH DESIGN This study was a retrospective secondary analysis of the data collected by the clinical core of RERC-SCI. Using the existing database information, specific variables pertaining to the Salzberg Scale were extracted to evaluate the psychometric properties of the scale. The extracted variables were the fifteen risk factors that composed the Salzberg Scale, demographic information of included subjects, and assessment dates for time point analysis in both acute care hospitalization and inpatient rehabilitation. The extracted variables were recorded at each subject assessment 17

37 The extracted data were then put into a separate database for lucid analysis of the particular psychometric properties of the scale. The psychometric properties evaluated were sensitivity, specificity, Receiver Operating Curve (ROC) Curve, and other predictive validity measures. These properties are explained in further detail in 4.4 DATA ANALYSES. Table 4 conveys the extracted variables used to organize the secondary analysis data. Table 4. Secondary Analysis Data Extracted Secondary Analysis Variables Categories Demographics Injury Secondary Analysis Variable Specifics Admission Date, Consent Date, Date of Injury, Age, Sex, Marital Status, Ethnicity, Height, Weight, Past Medical History (PMH) of Smoking, Rehabilitation Stage, Assessment Date, Current Residence Complete vs. Incomplete; American Spinal Injury Classification (ASIA) score, Medical Information Tobacco Use, Medical Comorbidities (cardiovascular, hematopoietic, respiratory, gastrointestinal, neurologic, immunological, psychiatric, integumentary), Number of Pressure Ulcers, Ambulation Status, Albumin Levels, Hematocrit Levels, Bladder and Bowl Management Pressure Ulcer Incidence Date 4.2 SUBJECT INCLUSION AND EXCLUSION CRITERIA Primary Study Subject Inclusion and Exclusion Criteria In the primary RERC on SCI study, subjects with new, traumatic SCI were recruited within hours after admission to UPMC Neurotrauma Centers. Subjects were eligible to be in the study if they met the following criteria: 1) Received acute medical and/or surgical treatment at UPMC hospitals 18

38 2) Received acute rehabilitation at Institution of Rehabilitation Research (IRR) South Side or UPMC Mercy after the acute medical and/or surgical treatment 3) Are 18 years and older 4) New, acute traumatic SCI (e.g. motor vehicle accident, fall, sports injury, gunshot wound) Subjects were to be automatically excluded from the study if they had any of the following: 1) Pre-existing diseases that would affect the inflammatory response to SCI (e.g. autoimmune or demyelinating diseases) 2) Previous SCI or other neurological dieses that affect the motor or sensory function of the subject In the RERC on SCI study, 104 subjects were enrolled from 2008 through At the end of the study, 48 subjects were still enrolled in the study, 13 subjects were out of the study due to death or lost to follow-up, and the remaining 43 subjects withdrew themselves from the study for reasons such as lack of interest or dislike of blood draws Secondary Study Subject Inclusion and Exclusion Criteria For the current analysis, 51 subjects of the original 104 subjects were included for this study. Subjects were eligible to participate in the current study if they met the following criteria: 1) Did not voluntarily withdraw from RERC-SCI study 2) Contain all information required for the Salzberg Scale component extraction and other analysis data (e.g., Braden Scale, ASIA score, ambulation status) Subjects who were categorized as withdrawn were not allowed to be used in the study as past medical records could not be accessed to extract certain information to complete the 19

39 Salzberg Scale assessment. Inaccessibility to data extraction was due to restrictions of the IRB. The items that needed to be extracted were albumin levels and hematocrit levels for each assessment date in order to have all variables of the Salzberg Scale accounted for in a score compilation. Subjects needed to have either albumin or total protein levels listed for the exact assessment date. Hematocrit levels also needed to be listed for the exact assessment date to maintain consistency and accurate data extraction. These values were not recorded for in the original database of the study. Access to this information could only be obtained through already consented patients or those who had passed away or had been lost to follow-up based on IRB specifications. 4.3 DATA COLLECTION Data extraction from the original RERC-SCI study was a crucial component to the success of the compilation of Salzberg Scale scores. Two databases were used for the extraction of the data. The first was the original database used to hold the primary RERC-SCI study. This held the majority of the sought after variables. Categorical titles of information held in this database can be referred to in Table 3. The second database of information was the computerized medical records pertaining to each subject. Albumin and/or total protein levels and hematocrit levels for each included subject were found in these records. A secondary analysis spreadsheet was used to organize and input data specific to the current study for statistical analysis. This spreadsheet held categories for each risk factor of the Salzberg Scale, assessment dates and phase specifications (acute hospitalization or inpatient rehabilitation), ASIA scores, and demographic information. Exact categorical topics in the secondary analysis spreadsheet can be referred to in Table 4. 20

40 To begin data extraction for the secondary analysis database, operational definitions of the Salzberg Scale factors were first studied in detail in order to look for exact variables in the original database and computerized medical records of subjects. Operational definitions ranged from International Classification of Disease (ICD) definitions, quantitative measurements, to objective observational details. The objective observational details pertained to three of the fifteen risk factors. The three factors were level of activity, mobility, and urine incontinence or constantly moist. These factors were coded based on the expert opinion of the two nurses who assessed each subject in acute hospital and inpatient rehabilitation settings. Each of the nurses reviewed extensive medical record notation and assessment notes of each subject pertaining to the three risk factors. In order to eliminate bias or skewed coding of the three risk factors, the nurses of the primary study were given practice evaluations or scorings to compare the assessments. Out of 48 variables from 12 subjects, the nurses overlapped with identical scoring 45 times (94%) showing strong consistency and similar coding methodology. Within this population, the limiting factor or variable to determine if the subject would be included was albumin or total protein measurement availability. This was the first risk factor that was assessed to minimize obsolete data collection. For each qualified subject, albumin levels were taken from the medical records. This narrowed our search to 51 subjects that allowed calculation of complete Salzberg Scale scores. Based on the operational definition as well as measured value, a coded value was assigned based on the Salzberg Scale. This process was repeated for all fifteen risk factors for each assessment date. The only varying step was from where the data were taken in the second step of the process (See Figure 1 Data Collection Process for flowchart of data collection process). 21

41 Review Salzberg Risk Factor and Operation Definition (OP) Extract RERC- SCI or medical record data based on Salzberg OP and input in secondary analysis database Assign Salzberg Coded Value based on RERC- SCI in Salzberg Database Add Coded Values of all fifteen risk factors for a complete Salzberg Score Figure 1 Data Collection Process Whether data were extracted from the RERC-SCI primary database of computerized medical records, variables were based on exact Salzberg operational definitions. Once each variable had been collected, all variables are summed together for a compiled Salzberg Scale Score. This was the number that was used in statistical analysis. Although each subject had complete Salzberg Scale scores for individual assessment dates, many subjects had missing data for Salzberg Scale score calculations over a large range of visits. A forward-carrying imputation method was initially attempted to account for each subject s missing scores. Based on large scale missing data of multiple subjects, an imputation method could not be used, as there would be too great of an assumption factor made when analyzing the dataset. Alternatively, a methodology evaluating the short term and long term use of the Salzberg Scale was established to eliminate possibility of skewed data. 22

42 The available data was analyzed two ways to assess the predictive validity of the scale for this study population (see Figure 2 ). The first analysis looked at the short-term risk assessment. This grouping of the data, also known as next visit prediction group, would only include subjects who had a Salzberg Scale score followed by a study follow up for assessment of pressure ulcer development. When the subjects were scored in acute hospitalization, then the subject s pressure ulcer assessment would range two to three days after the initial Salzberg Scale score compilation, as subjects were assessed three times a week as part of the primary study protocol. If the subject were scored in inpatient rehabilitation, then the subject s pressure ulcer assessment would be five to seven days after the initial Salzberg Score compilation as subjects were assessed one time per week. This subject group captures the predictive validity of the Salzberg Scale when administered in a short-term, frequent method. Isolation of a direct time point for pressure ulcer formation could be highly correlated to a specific range of Salzberg Scores, as the obtained score of the initial visit would likely estimate the development of the PrU upon the following visit. Although only 28 subjects were included in this group, 55 pairings were gathered from these subjects. Two subjects were included in both the acute and inpatient settings whereas the other subjects were in either acute hospitalization on inpatient rehabilitation settings. Individual subjects contributed a range of one to six observation pairings to the 55 pairings used in statistical analysis as long as the assessment pairings fit the next visit prediction group criteria. (See Table 5) 23

43 n=11 Figure 2 Subject Selection Flow Chart 24